In designing high performance integrated circuits, the need to transfer addresses and data across a bus at high speed is a critical consideration. This is especially true in applications where a memory and a high performance state machine are being integrated into a single chip. One such instance is the bus between a display controller and a frame buffer memory. In this case, substantial amounts of data, and the corresponding addresses, must be transferred between the graphics controller and the frame buffer at rates high enough to support display refresh and update, and other processing operations, such as filtering. As display systems with increased resolution and bit depths are developed, the rate at which data must be transferred between the controller and the frame buffer consequently increases. While some of the necessary bandwidth can be achieved by using wider buses, improvement in the speed at which data is transferred over the individual bus lines is still required.
The lines of a typical on-chip bus or PC board bus are long, thin conductors which extend relatively substantial lengths across the face of an insulating substrate, the substrate spacing each conductor from a ground plane and form other signal wires. As a result of this configuration, each line presents a significant capacitance which must be charged or discharged by a bus driver or similar circuit during data transmission. The result is substantial power consumption, particularly when a CMOS or TTL bus is driven between positive and negative power supply rails.
The power consumption resulting from parasitic bus line capacitance increases directly with the data transmission rate across the bus, since P=CV.sup.2 f, where P is the power loss through each conductor, V is the voltage applied, C is the capacitance of the conductor, and f is the frequency at which the conductor is charged/discharged. It should also be noted that some additional small power consumption results from the resistance of each bus line.
It is possible to reduce power consumption by reducing the capacitance of the bus lines themselves. This option, however, requires that the fabrication process for chips and for circuit boards be modified; a change in process to reduce line capacitance is expensive and may adversely effect the fabrication of other circuitry on chips and boards. Another option is to reduce the frequency at which data is transferred across the bus. Assuming that the width of the bus is not increased, this option simply trades off system performance for power reduction, an option which usually is not viable in the design and implementation of high performance circuits.
Thus, the need has arisen for improved circuits, systems and methods for transferring data and/or addresses across the lines of a bus at high rates. Such circuits, systems and methods should advantageously minimize power consumption and the problems attendant therewith. In particular, such circuits, systems and methods should be applicable to high performance integrated circuit applications, such as display controllers and semiconductor memories. Finally, such circuits, systems and methods should require neither expensive and complicated changes to the chip fabrication process nor a reduction in system performance for implementation.